Claude Code Leaks Users' Email Addresses to Websites
A privacy issue in Claude Code unintentionally leaked users' personal and organizational email addresses to visited websites via context injection. Anthropic quickly responded by releasing version 2.1.317, which includes a built-in module named `cc-plugin-privacy` to block this behavior. This bug raised significant security and privacy concerns, as developer credentials and personal identifiers were exposed to third-party servers without user consent. The incident highlights the risks of automated coding assistants inadvertently mishandling sensitive user data within their operational context. Affected users reported an increase in personalized spam emails after using the tool. Anthropic's mitigation can be optionally disabled or managed via the `/plugins` command for users running version 2.1.317 or later.
## BACKGROUND
Claude Code is a terminal-based AI coding assistant developed by Anthropic that helps developers write and debug code. Context injection is a mechanism where an AI application automatically passes relevant state or environment details into the model's working context to assist with tasks.